AI Agent Hub
Back to plugins
🖥️

dsh-collapsible-models

Client Updated 2026.08.23

Run the following command in DeepSeek Harness:

dsh plugin install ZhenXifu/dsh-collapsible-models

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install ZhenXifu/dsh-collapsible-models in your terminal, or visit https://github.com/ZhenXifu/dsh-collapsible-models to install manually; restart the Harness client after installation for the collapsible picker to take effect.

About this plugin

With several model providers configured, the composer model picker turns into a scroll marathon. dsh-collapsible-models does exactly one thing well: it adds a collapsible triangle to every provider row in the dropdown, so you can fold or unfold each group with a single click and turn a flat wall of models into a tidy two-level list.

Under the hood the plugin registers a shadow entry at priority -1 over the stock picker while leaving the data layer completely untouched. Model catalog, current selection, reasoning-effort handling, loading states, and toast notifications all behave exactly as they do in the native UI. Collapse state lives only for the duration of the open dropdown and is never persisted. Group headers remain real, focusable elements, so keyboard navigation with Tab and Enter works unchanged, and aria-expanded plus accessible labels follow the active locale in both Chinese and English.

This is for anyone who has wired up multiple providers — OpenAI, Anthropic, a local Ollama instance, and friends — and ends up scrolling past dozens of models just to find the one they need. Install it, restart the harness once, and the picker becomes a clean provider-first, model-second flow without touching any existing configuration.

Use Cases

  • Managing multiple providers with an overwhelming model dropdown
  • Switching between providers quickly by collapsing irrelevant groups
  • Keyboard-first workflow that needs focusable group headers

Best For

  • Harness users with three or more model providers configured
  • Users who prefer a compact picker over scrolling through dozens of models
  • Developers who rely on keyboard navigation and screen readers